Server management method and system based on LightGBM model

Abstract

The invention discloses a server management method and system based on a LightGBM model, and relates to the technical field of server management.After health states of all servers are analyzed based on the LightGBM model, all the servers are classified into a first-class set, a second-class set and a third-class set; performing primary optimization adjustment on the monitoring frequencies of all the servers in the second-class set according to the health state analysis result, monitoring each server at the corresponding monitoring frequency, and performing dynamic monitoring frequency adjustment on the servers in the first-class set according to the adjustment table, and the servers in the second-class set perform secondary optimization adjustment on the monitoring frequency in combination with the real-time operation state. The management system divides the servers into a first-class set, a second-class set and a third-class set, so that each class of servers has a corresponding management strategy, and optimization can be performed according to health states and real-time operation data of the servers by dynamically adjusting the monitoring frequency.

Description

Technical Field

[0001] The present invention relates to the field of server management technology, and in particular to a server management method and system based on a LightGBM model. Background Art

[0002] A server is a type of computer system that is dedicated to providing data, services, and resources to other computers (clients) or devices. They play a key role in modern information technology infrastructure and are widely used in enterprises, data centers, cloud computing, and various online services. Servers usually have high performance, high availability, and scalability to meet the needs of continuous operation and processing a large number of requests.

[0003] The prior art has the following defects:

[0004] Management systems often fail to effectively classify servers, making it impossible to develop specific management strategies for servers in different health states, which may lead to waste of resources or premature retirement of equipment. Management systems also lack flexibility in monitoring frequency, often using a fixed monitoring frequency, which may result in missing important operating status changes or fault warnings.

[0005] Based on this, the present invention proposes a server management method and system based on the LightGBM model, which divides servers into one, two and three categories, so that each type of server has a corresponding management strategy. By dynamically adjusting the monitoring frequency, it can be optimized according to the health status and real-time operation data of the server, ensuring more timely discovery of potential problems and improving fault warning capabilities. Summary of the invention

[0006] The purpose of the present invention is to provide a server management method and system based on the LightGBM model to solve the shortcomings of the background technology.

[0007] In order to achieve the above object, the present invention provides the following technical solution: a server management method based on the LightGBM model, the management method comprising the following steps:

[0008] When all servers purchased by the enterprise are in trial operation, the management system obtains the trial operation data of all servers, analyzes the health status of all servers based on the LightGBM model, and classifies all servers into one-category set, two-category set, and three-category set;

[0009] Generate corresponding management policies for all servers in the third category set, and optimize the monitoring frequency of all servers in the second category set based on the health status analysis results;

[0010] During the use of all servers, the management system monitors each server at the corresponding monitoring frequency to obtain the real-time operating status of the server. For servers in the first category, the monitoring frequency is dynamically adjusted according to the adjustment table. For servers in the second category, the monitoring frequency is secondary optimized and adjusted in combination with the real-time operating status. For servers in the third category, the corresponding management strategy is generated based on the real-time operating status.

[0011] In a preferred embodiment, when all servers purchased by the enterprise are in trial operation, the management system obtains the trial operation data of all servers and analyzes the health status of all servers based on the LightGBM model, including the following steps:

[0012] Obtain the mechanical parameters, electrical parameters, and network parameters of the server;

[0013] Mechanical parameters include the temperature rise rate, vibration amplitude and fan noise decibel during the trial run; electrical parameters include power factor, current load and power supply unit efficiency during the trial run; network parameters include bandwidth utilization, delay rate, packet loss rate and network transmission error rate;

[0014] The mechanical parameters, electrical parameters, and network parameters are substituted into the pre-trained LightGBM model, and the LightGBM model generates a health status score for the server.

[0015] In a preferred embodiment, during the use of all servers, the management system monitors each server at a corresponding monitoring frequency to obtain the real-time operating status of the server, including the following steps:

[0016] During the use of all servers, the servers in the first set are monitored at the initial monitoring frequency, the servers in the second set are monitored at the once-adjusted monitoring frequency, and the servers in the third set are monitored at twice the initial monitoring frequency;

[0017] Obtain the bandwidth utilization, multi-region temperature difference factor and voltage fluctuation amplitude during the operation of the server; comprehensively calculate the bandwidth utilization, multi-region temperature difference factor and voltage fluctuation amplitude to obtain the operation status coefficient, the expression is: , where is the operating state coefficient, is the bandwidth utilization, is the multi-region temperature difference factor, is the voltage fluctuation amplitude, , , are the bandwidth utilization, multi-region temperature difference factor and voltage fluctuation amplitude adjustment coefficient, respectively, and , , Both are greater than 0. The larger the operating status coefficient is, the better the operating status of the server is during use.

[0018] In a preferred embodiment, all servers are classified into a first-class set, a second-class set, and a third-class set, including the following steps:

[0019] Analyze the health status of all servers based on the LightGBM model, and output the health status scores of all servers. The larger the health status score, the better the health status of the server. Compare the obtained health status score with the preset first score threshold and the second score threshold.

[0020] If the health status score is greater than or equal to the second scoring threshold, the health status of the server is analyzed to be excellent, and the server is classified into the first category. If the health status score is less than the second scoring threshold, and the health status score is greater than or equal to the first scoring threshold, the health status of the server is analyzed to be medium, and the server is classified into the second category. If the health status score is less than the first scoring threshold, the health status of the server is analyzed to be poor, and the server is classified into the third category.

[0021] In a preferred embodiment, the monitoring frequency of all servers in the second set is optimized and adjusted according to the health status analysis result, including the following steps:

[0022] After obtaining the health status scores of all servers in the second set, the initial monitoring frequency of the servers in the second set is optimized and adjusted according to the health status scores. That is, the smaller the health status score of the server, the more it is necessary to increase the initial monitoring frequency of the server. The expression of the adjustment algorithm is: , where is the monitoring frequency after one adjustment, is the initial monitoring frequency, is the maximum value of health status score, Score the current health of the server.

[0023] In a preferred embodiment, the calculation logic of the bandwidth utilization is: obtaining the total bandwidth of the network, obtaining the actual bandwidth used by the server, and dividing the actual bandwidth used by the total bandwidth to obtain the bandwidth utilization;

[0024] The calculation logic of the multi-region temperature difference factor is as follows: divide the server into multiple monitoring areas, obtain the number of components in each area, calculate the temperature difference between multiple components in one monitoring area, sum the temperature differences between multiple components to obtain the total temperature difference value of the monitoring area, and sum the total temperature differences of multiple monitoring areas to obtain the multi-region temperature difference factor;

[0025] The calculation logic of the voltage fluctuation amplitude is: obtain the voltage values ​​of the server at multiple time points, and calculate the voltage fluctuation amplitude based on the voltage values ​​at multiple time points. The expression is: , where is the voltage fluctuation amplitude, is the number of time points, For the The voltage value obtained at a time point, is the average voltage value.

[0026] In a preferred embodiment, for servers in a class set, dynamic monitoring frequency adjustment is performed according to an adjustment table, including the following steps:

[0027] After obtaining the operating status coefficients of all servers in a class set, the operating status coefficients are compared with a preset first coefficient threshold and a second coefficient threshold, the first coefficient threshold is used to determine whether it is necessary to increase the monitoring frequency of the server, and the second coefficient threshold is used to determine whether it is necessary to reduce the monitoring frequency of the server;

[0028] If the operating status coefficient is greater than or equal to the first coefficient threshold, and the operating status coefficient is less than or equal to the second coefficient threshold, it is determined that there is no need to adjust the monitoring frequency of a class of collection servers; if the operating status coefficient is less than the first coefficient threshold, it is determined that the monitoring frequency of a class of collection servers needs to be increased; if the operating status coefficient is greater than the second coefficient threshold, it is determined that the monitoring frequency of a class of collection servers needs to be reduced;

[0029] When it is determined that the monitoring frequency of a class of collection servers needs to be increased or decreased, dynamic monitoring frequency adjustment is performed according to the adjustment table.

[0030] In a preferred embodiment, for the servers in the second set, the monitoring frequency is optimized and adjusted twice in combination with the real-time operation status, including the following steps:

[0031] Get the operating status coefficients of all servers in the second set. If the operating status coefficient of the server is less than or equal to the second coefficient threshold, the monitoring frequency of the server is not adjusted for secondary optimization. If the operating status coefficient of the server is greater than the second coefficient threshold, the monitoring frequency is adjusted by the operating status coefficient. The adjustment expression is: , where is the monitoring frequency after secondary adjustment, is the monitoring frequency after one adjustment, is the operating status coefficient.

[0032] In a preferred embodiment, generating corresponding management policies for all servers in the three types of sets includes the following steps:

[0033] After the management system obtains the information of all servers in the three categories, it marks all servers with a recheck mark and sends it to the administrator;

[0034] After all servers marked with re-inspection marks have been inspected and repaired, the health status of all servers that have completed inspection and repair is analyzed again based on the LightGBM model;

[0035] If the server that has been overhauled is classified as a Class II set or a Class I set, the re-inspection mark of the server is deleted and the server is put into use;

[0036] If the server that has been repaired is still classified as a Class III collection, the server will be deleted from the system library and will no longer be used.

[0037] The server management system based on the LightGBM model includes a set partitioning module, an optimization management module, and a secondary optimization module;

[0038] Set division module: When all servers purchased by the enterprise are in trial operation, the trial operation data of all servers are obtained, and after analyzing the health status of all servers based on the LightGBM model, all servers are classified into first-class set, second-class set and third-class set;

[0039] Optimization management module: Generates corresponding management policies for all servers in the third category set, and optimizes the monitoring frequency of all servers in the second category set based on the health status analysis results;

[0040] Secondary optimization module: During the use of all servers, each server is monitored at the corresponding monitoring frequency to obtain the real-time operating status of the server. For servers in the first category, the monitoring frequency is dynamically adjusted according to the adjustment table. For servers in the second category, the monitoring frequency is secondarily optimized and adjusted in combination with the real-time operating status. For servers in the third category, corresponding management strategies are generated based on the real-time operating status.

[0041] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0042] After analyzing the health status of all servers based on the LightGBM model, the present invention classifies all servers into one set, two sets and three sets, generates corresponding management strategies for all servers in the three sets, optimizes and adjusts the monitoring frequency of all servers in the two sets according to the health status analysis results, monitors each server at the corresponding monitoring frequency, obtains the real-time operation status of the server, and dynamically adjusts the monitoring frequency for servers in the one set according to the adjustment table. For servers in the two sets, the monitoring frequency is optimized twice in combination with the real-time operation status. The management system divides servers into one, two and three sets, so that each type of server has a corresponding management strategy. By dynamically adjusting the monitoring frequency, it can be optimized according to the health status and real-time operation data of the server, ensuring that potential problems are discovered more timely and improving fault warning capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0044] Figure 1 The present invention is a flow chart of the method. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] Example 1: Please refer to Figure 1 As shown, the server management method based on the LightGBM model described in this embodiment includes the following steps:

[0047] When all servers purchased by the enterprise are in trial operation, the management system obtains the trial operation data of all servers, and after analyzing the health status of all servers based on the LightGBM model, all servers are classified into one set, two sets and three sets, and corresponding management policies are generated for all servers in the three sets. The monitoring frequency of all servers in the two sets is optimized and adjusted according to the health status analysis results. During the use of all servers, the management system monitors each server at the corresponding monitoring frequency to obtain the real-time operation status of the server. For servers in the one set, the monitoring frequency is dynamically adjusted according to the adjustment table. For servers in the two sets, the monitoring frequency is optimized and adjusted for the second time in combination with the real-time operation status. For servers in the three sets, corresponding management policies are generated based on the real-time operation status.

[0048] After analyzing the health status of all servers based on the LightGBM model, this application classifies all servers into one set, two sets, and three sets, generates corresponding management policies for all servers in the three sets, optimizes and adjusts the monitoring frequency of all servers in the second set based on the health status analysis results, monitors each server at the corresponding monitoring frequency, obtains the real-time operating status of the server, and dynamically adjusts the monitoring frequency for servers in the first set according to the adjustment table. For servers in the second set, the monitoring frequency is optimized and adjusted twice based on the real-time operating status. The management system divides servers into one, two, and three sets, so that each type of server has a corresponding management strategy. By dynamically adjusting the monitoring frequency, it can be optimized according to the health status and real-time operating data of the server, ensuring that potential problems are discovered more promptly and improving fault warning capabilities.

[0049] Embodiment 2: When all servers purchased by the enterprise are in trial operation, the management system obtains the trial operation data of all servers and analyzes the health status of all servers based on the LightGBM model, including the following steps:

[0050] Obtain the mechanical parameters, electrical parameters, and network parameters of the server;

[0051] Mechanical parameters include the temperature rise rate, vibration amplitude, and fan noise decibels during the test run;

[0052] Power parameters include power factor, current load, and power supply unit efficiency during the trial operation;

[0053] Network parameters include bandwidth utilization, delay rate, packet loss rate, and network transmission error rate;

[0054] The mechanical parameters, electrical parameters, and network parameters are substituted into the pre-trained LightGBM model, and the LightGBM model generates a health status score for the server.

[0055] The pre-training of the LightGBM model includes the following steps:

[0056] After collecting the temperature rise rate, vibration amplitude, fan noise decibel, power factor, current load, power supply unit efficiency, bandwidth utilization, delay rate, packet loss rate and network transmission error rate of several servers, we put them into the linear regression algorithm for analysis. The algorithm is expressed as follows: , where Score your health status. are variables (various parameters), is the regression coefficient of each variable, and m represents the number of parameters;

[0057] In this embodiment, the regression coefficient The value of is set according to the effect of the variable on the health status of the server (such as increasing the health status or decreasing the health status). When the variable causes the health status of the server to increase, the regression coefficient , when the variable causes a decrease in server health, the regression coefficient , specifically:

[0058] Regression coefficient greater than 0: regression coefficient corresponding to power factor, power supply unit efficiency, and bandwidth utilization;

[0059] Regression coefficient is less than 0: regression coefficient corresponding to temperature rise rate, vibration amplitude, fan noise decibel, current load, delay rate, packet loss rate and network transmission error rate;

[0060] When the present invention is used, the logical factors of the health status score are composed of: taking the influence of variables on the health status of the server as an example, the first is the indicator, that is, the factor that causes the change in the health status of the server (the present invention refers to the influence of variables on the health status of the server); the second is the weight of these indicators, that is, the proportion of each variable when it is generated; the third is the operation equation, that is, what kind of mathematical operation process is used to obtain the result, and the health status score is obtained by calculating the indicators with their respective weights through the operation equation.

[0061] The variables obtained from the sample were transformed and processed into data language recognized by computer software; secondly, these evaluation factors were analyzed by Logistic regression using SPSS software to screen out factors and their weights that are significantly correlated with the results; thirdly, the evaluation factors and weights were brought into the Logistic regression equation for calculation to obtain the results, which are as follows:

[0062] First, ensure the integrity of the variables, handle missing values ​​and outliers, and convert the data into a format that SPSS software can recognize. Usually, the data is stored in csv, xlsx and other formats, and then imported into SPSS. Open SPSS software, import the processed data file, and transform the variables as needed. For example, for continuous variables, standardize or normalize them. Select the "Analyze" menu, then select the "Binary Logistic" option under "Regression". In the dialog box, add the dependent variable (outcome) and independent variable (variable) to the corresponding boxes. SPSS will fit the Logistic regression model based on the selected variables. In the output results, you will see The coefficients, standard errors, p-values ​​and other information of the model. Check the coefficients and p-values ​​in the output results to determine which variables are significantly correlated with the results. Usually, a p-value less than 0.05 is considered significant. While fitting the model, use variable selection methods, such as stepwise regression, to help screen the most relevant factors. According to the coefficients of the Logistic regression model, the size of the coefficients reflects the degree of influence of each factor on the result, and the positive and negative signs of the coefficients indicate the direction of the influence. After obtaining the significant factors and their coefficients, the Logistic regression equation is obtained, which is used to calculate the probability of each sample, and then predict the results to complete the pre-training of the LightGBM model.

[0063] Classifying all servers into first-category sets, second-category sets, and third-category sets includes the following steps:

[0064] Analyze the health status of all servers based on the LightGBM model, and output the health status scores of all servers. The larger the health status score, the better the health status of the server. Compare the obtained health status score with the preset first score threshold and the second score threshold. The first score threshold is less than the second score threshold.

[0065] If the health status score is greater than or equal to the second scoring threshold, the health status of the server is analyzed to be excellent, and the server is classified into the first category. If the health status score is less than the second scoring threshold, and the health status score is greater than or equal to the first scoring threshold, the health status of the server is analyzed to be medium, and the server is classified into the second category. If the health status score is less than the first scoring threshold, the health status of the server is analyzed to be poor, and the server is classified into the third category.

[0066] In this application, dividing the server health status scores into three sets (excellent, medium, and poor) has the following benefits:

[0067] Clear management strategy: According to the different health status, formulate corresponding management strategies for each collection to make resource allocation more effective. For example, the best class servers can be kept in normal use, the medium class servers need regular monitoring, and the poor class servers should be given priority for repair or replacement.

[0068] Improve operation and maintenance efficiency: Through classification, you can optimize the monitoring frequency and maintenance plan, focus on servers with poor health, and reduce unnecessary monitoring work.

[0069] Preventive maintenance: Timely identification of servers in moderate health status helps in preventive maintenance, reduces the risk of potential failures, and extends the life of the equipment.

[0070] Data-driven decision making: By categorizing health status scores, better data analysis can be performed to identify key factors that affect health status, thereby making more informed choices in future purchasing and management decisions.

[0071] Reduce operating costs: By effectively managing servers in different health states, you can reduce failure rates, downtime and repair costs, and improve overall operational efficiency.

[0072] Enhanced security: Timely detection of servers in poor health helps avoid data loss or service interruptions and improves system security and stability.

[0073] Generate corresponding management policies for all servers in the three types of collections, including the following steps:

[0074] All servers in the three types of collections are in poor health, which may cause the following problems for the enterprise if they are put into use:

[0075] System failure: Servers in poor health are more susceptible to hardware failures, which can cause system crashes or downtime, affecting business continuity.

[0076] Data loss: Hardware problems can lead to data corruption or loss, which can be costly, especially if the server is not backed up regularly.

[0077] Performance degradation: Low operating efficiency and long response time affect user experience and lead to customer loss.

[0078] Increased maintenance costs: Frequent failures require constant maintenance and repair, which increases operation and maintenance costs and human resource consumption.

[0079] Security risks: Old or unstable servers may have security vulnerabilities and be vulnerable to cyber attacks, leading to data leaks or security incidents.

[0080] Impact on corporate reputation: Unstable services may affect customer satisfaction and damage corporate reputation and brand image.

[0081] Waste of resources: Continuing to use a server in poor health wastes resources and budget, and should be prioritized for replacement or upgrading.

[0082] Compliance risk: In some industries, using non-compliant equipment may violate regulatory requirements, leading to legal and financial risks.

[0083] Therefore, the management strategy for servers in the three types of collections is:

[0084] After the management system obtains the information of all servers in the three-category collection, it marks all servers with a re-inspection mark and sends it to the administrator. The administrator sends all servers marked with the re-inspection mark to the maintenance station for inspection and maintenance. After the inspection and maintenance of all servers marked with the re-inspection mark are completed, the health status of all servers that have completed inspection and maintenance is re-analyzed based on the LightGBM model. If the server that has completed maintenance is classified as a second-category collection or a first-category collection, the re-inspection mark of the server is deleted and put into use. If the server that has completed maintenance is still classified as a third-category collection, the server is deleted from the system library and no longer used.

[0085] When a server in a Class III collection is reclassified as a Class II collection or a Class I collection, the Class III collection identifier is retained.

[0086] Through this management process, enterprises can effectively monitor and maintain the health of servers, ensure that the equipment put into use meets the standards, thereby reducing the risk of failure and operating costs. For servers with poor health, timely detection and processing can avoid potential problems and ensure the stable operation of the enterprise.

[0087] According to the health status analysis results, the monitoring frequency of all servers in the second category set is optimized and adjusted, including the following steps:

[0088] The health status of all servers in the second category is medium, that is, these servers can continue to run, but their performance has declined. Therefore, the monitoring frequency of all servers in the second category needs to be increased;

[0089] In this application, the initial monitoring frequency for all servers is a general monitoring frequency, that is, the initial monitoring frequency can both ensure the monitoring effect of the server and reduce the operating burden and energy consumption of the management system;

[0090] The servers in the first category are monitored at the initial monitoring frequency;

[0091] After obtaining the health status scores of all servers in the second set, the initial monitoring frequency of the servers in the second set is optimized and adjusted according to the health status scores. That is, the smaller the health status score of the server, the more it is necessary to increase the initial monitoring frequency of the server. The expression of the adjustment algorithm is: , where is the monitoring frequency after one adjustment, is the initial monitoring frequency, is the maximum value of health status score, Score the current health of the server.

[0092] During the use of all servers, the management system monitors each server at a corresponding monitoring frequency to obtain the real-time operating status of the server, including the following steps:

[0093] During the use of all servers, the servers in the first set are monitored at the initial monitoring frequency, the servers in the second set are monitored at the once-adjusted monitoring frequency, and the servers in the third set are monitored at twice the initial monitoring frequency;

[0094] Obtain bandwidth utilization, multi-region temperature difference factors, and voltage fluctuation amplitude during server operation;

[0095] The bandwidth utilization, multi-region temperature difference factor and voltage fluctuation amplitude are calculated comprehensively to obtain the operating state coefficient, and the expression is: , where is the operating state coefficient, is the bandwidth utilization, is the multi-region temperature difference factor, is the voltage fluctuation amplitude, , , are the bandwidth utilization, multi-region temperature difference factor and voltage fluctuation amplitude adjustment coefficient, respectively, and , , Both are greater than 0. The larger the operating status coefficient is, the better the operating status of the server is during use.

[0096] The calculation logic of bandwidth utilization is as follows: obtain the total bandwidth of the network, obtain the actual bandwidth used by the server, and divide the actual bandwidth used by the total bandwidth to obtain the bandwidth utilization. The higher the bandwidth utilization, the better the running status of the server.

[0097] The calculation logic of the multi-area temperature difference factor is: divide the server into multiple monitoring areas, obtain the number of components in each area, calculate the temperature difference between multiple components in one monitoring area, sum the temperature difference between multiple components to obtain the total temperature difference value of the monitoring area, sum the total temperature difference values ​​of multiple monitoring areas to obtain the multi-area temperature difference factor. The larger the multi-area temperature difference factor, the greater the temperature difference inside the server, which indicates that the server's operating status is worse.

[0098] The calculation logic of the voltage fluctuation amplitude is: obtain the voltage values ​​of the server at multiple time points, and calculate the voltage fluctuation amplitude based on the voltage values ​​at multiple time points. The expression is: , where is the voltage fluctuation amplitude, is the number of time points, For the The voltage value obtained at a time point, is the average voltage value. The larger the voltage fluctuation, the more unstable the voltage of the server is and the worse the operating status is.

[0099] For servers in a class set, dynamic monitoring frequency adjustment is performed according to the adjustment table, including the following steps:

[0100] After obtaining the operating status coefficients of all servers in a class set, the operating status coefficients are compared with a preset first coefficient threshold and a second coefficient threshold, the first coefficient threshold is used to determine whether it is necessary to increase the monitoring frequency of the server, and the second coefficient threshold is used to determine whether it is necessary to reduce the monitoring frequency of the server;

[0101] If the operating status coefficient is greater than or equal to the first coefficient threshold, and the operating status coefficient is less than or equal to the second coefficient threshold, it is determined that there is no need to adjust the monitoring frequency of a class of collection servers; if the operating status coefficient is less than the first coefficient threshold, it is determined that the monitoring frequency of a class of collection servers needs to be increased; if the operating status coefficient is greater than the second coefficient threshold, it is determined that the monitoring frequency of a class of collection servers needs to be reduced;

[0102] When it is determined that the monitoring frequency of a class of collection servers needs to be increased or decreased, the dynamic monitoring frequency adjustment is performed according to the adjustment table, and the adjustment table is shown in Table 1:

[0103] Table 1 Adjustment table

[0104]

[0105] In Table 1, is the adjusted monitoring frequency, is the monitoring frequency before adjustment, is the first coefficient threshold, is the second coefficient threshold, To reduce the gradient threshold A, To reduce the gradient threshold B, To reduce the gradient threshold C, To reduce the gradient threshold D, To increase the gradient threshold A, To increase the gradient threshold B, To increase the gradient threshold C, In order to increase the gradient threshold D, for a class of servers in a set, the smaller the operating status coefficient, the more the monitoring frequency increases, that is, the monitoring intensity needs to be strengthened; the larger the operating status coefficient, the more the monitoring frequency decreases, that is, the monitoring intensity needs to be slowed down; when the operating status coefficient is between the first coefficient threshold and the second coefficient threshold, there is no need to dynamically adjust the monitoring frequency.

[0106] For servers in the second category, the monitoring frequency is optimized and adjusted again based on the real-time operation status, including the following steps:

[0107] Get the operating status coefficients of all servers in the second set. If the operating status coefficient of the server is less than or equal to the second coefficient threshold, the monitoring frequency of the server is not adjusted for secondary optimization. If the operating status coefficient of the server is greater than the second coefficient threshold, the monitoring frequency is adjusted by the operating status coefficient. The adjustment expression is: , where is the monitoring frequency after secondary adjustment, is the monitoring frequency after one adjustment, is the operating status coefficient.

[0108] For servers in the three types of sets, corresponding management policies are generated based on the real-time operating status, including the following steps:

[0109] The operating status coefficients of the servers in the three types of sets are obtained. If the operating status coefficient of the server is less than a first coefficient threshold, the server is marked as unusable and a warning signal is sent to an administrator.

[0110] Since the servers in the third category set are in poor health during the trial operation and are still unstable during use after maintenance, it is not recommended to continue using them in order to ensure the stable operation of the enterprise.

[0111] Embodiment 3: The server management system based on the LightGBM model described in this embodiment includes a set partitioning module, an optimization management module, and a secondary optimization module;

[0112] Set division module: When all servers purchased by the enterprise are in trial operation, the trial operation data of all servers are obtained, and after analyzing the health status of all servers based on the LightGBM model, all servers are classified into first-class sets, second-class sets, and third-class sets. The set division results are sent to the optimization management module;

[0113] Optimization management module: Generates corresponding management policies for all servers in the third set, optimizes and adjusts the monitoring frequency of all servers in the second set based on the health status analysis results, and sends the corresponding monitoring frequencies of all servers to the secondary optimization module;

[0114] Secondary optimization module: During the use of all servers, each server is monitored at the corresponding monitoring frequency to obtain the real-time operating status of the server. For servers in the first category, the monitoring frequency is dynamically adjusted according to the adjustment table. For servers in the second category, the monitoring frequency is secondarily optimized and adjusted in combination with the real-time operating status. For servers in the third category, corresponding management strategies are generated based on the real-time operating status.

[0115] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0116] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0117] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well.

Claims

1. A server management method based on the LightGBM model, characterized by: The management approach includes the following steps: When all servers purchased by the enterprise are in trial operation, the management system obtains the trial operation data of all servers, analyzes the health status of all servers based on the LightGBM model, and classifies all servers into one-category set, two-category set, and three-category set; Generate corresponding management policies for all servers in the third category set, and optimize the monitoring frequency of all servers in the second category set based on the health status analysis results; During the use of all servers, the management system monitors each server at the corresponding monitoring frequency to obtain the real-time operating status of the server. For servers in the first category, the monitoring frequency is dynamically adjusted according to the adjustment table. For servers in the second category, the monitoring frequency is secondary optimized and adjusted in combination with the real-time operating status. For servers in the third category, the corresponding management strategy is generated based on the real-time operating status.

2. The server management method based on the LightGBM model according to claim 1, characterized in that: When all servers purchased by the enterprise are in trial operation, the management system obtains the trial operation data of all servers and analyzes the health status of all servers based on the LightGBM model, including the following steps: Obtain the mechanical parameters, electrical parameters, and network parameters of the server; Mechanical parameters include the temperature rise rate, vibration amplitude and fan noise decibel during the trial run; electrical parameters include power factor, current load and power supply unit efficiency during the trial run; network parameters include bandwidth utilization, delay rate, packet loss rate and network transmission error rate; The mechanical parameters, electrical parameters, and network parameters are substituted into the pre-trained LightGBM model, and the LightGBM model generates a health status score for the server.

3. The server management method based on the LightGBM model according to claim 2 is characterized in that: During the use of all servers, the management system monitors each server at a corresponding monitoring frequency to obtain the real-time operating status of the server, including the following steps: During the use of all servers, the servers in the first set are monitored at the initial monitoring frequency, the servers in the second set are monitored at the once-adjusted monitoring frequency, and the servers in the third set are monitored at twice the initial monitoring frequency; Obtain the bandwidth utilization, multi-region temperature difference factor and voltage fluctuation amplitude during the operation of the server; comprehensively calculate the bandwidth utilization, multi-region temperature difference factor and voltage fluctuation amplitude to obtain the operation status coefficient, the expression is: , where is the operating state coefficient, is the bandwidth utilization, is the multi-region temperature difference factor, is the voltage fluctuation amplitude, , , are the bandwidth utilization, multi-region temperature difference factor and voltage fluctuation amplitude adjustment coefficient, respectively, and , , Both are greater than 0. The larger the operating status coefficient is, the better the operating status of the server is during use.

4. The server management method based on the LightGBM model according to claim 3 is characterized in that: Classifying all servers into first-category sets, second-category sets, and third-category sets includes the following steps: Analyze the health status of all servers based on the LightGBM model, and output the health status scores of all servers. The larger the health status score, the better the health status of the server. Compare the obtained health status score with the preset first score threshold and the second score threshold. If the health status score is greater than or equal to the second scoring threshold, the health status of the server is analyzed to be excellent, and the server is classified into the first category. If the health status score is less than the second scoring threshold, and the health status score is greater than or equal to the first scoring threshold, the health status of the server is analyzed to be medium, and the server is classified into the second category. If the health status score is less than the first scoring threshold, the health status of the server is analyzed to be poor, and the server is classified into the third category.

5. The server management method based on the LightGBM model according to claim 4 is characterized in that: According to the health status analysis results, the monitoring frequency of all servers in the second category set is optimized and adjusted, including the following steps: After obtaining the health status scores of all servers in the second set, the initial monitoring frequency of the servers in the second set is optimized and adjusted according to the health status scores. That is, the smaller the health status score of the server, the more it is necessary to increase the initial monitoring frequency of the server. The expression of the adjustment algorithm is: , where is the monitoring frequency after one adjustment, is the initial monitoring frequency, is the maximum value of health status score, Score the current health of the server.

6. The server management method based on the LightGBM model according to claim 5 is characterized in that: The calculation logic of bandwidth utilization is as follows: obtain the total bandwidth of the network, obtain the actual bandwidth used by the server, and divide the actual bandwidth used by the total bandwidth to obtain the bandwidth utilization; The calculation logic of the multi-region temperature difference factor is as follows: divide the server into multiple monitoring areas, obtain the number of components in each area, calculate the temperature difference between multiple components in one monitoring area, sum the temperature differences between multiple components to obtain the total temperature difference value of the monitoring area, and sum the total temperature differences of multiple monitoring areas to obtain the multi-region temperature difference factor; The calculation logic of the voltage fluctuation amplitude is: obtain the voltage values ​​of the server at multiple time points, and calculate the voltage fluctuation amplitude based on the voltage values ​​at multiple time points. The expression is: , where is the voltage fluctuation amplitude, is the number of time points, For the The voltage value obtained at a time point, is the average voltage value.

7. The server management method based on the LightGBM model according to claim 6 is characterized in that: For servers in a class set, dynamic monitoring frequency adjustment is performed according to the adjustment table, including the following steps: After obtaining the operating status coefficients of all servers in a class set, the operating status coefficients are compared with a preset first coefficient threshold and a second coefficient threshold, the first coefficient threshold is used to determine whether it is necessary to increase the monitoring frequency of the server, and the second coefficient threshold is used to determine whether it is necessary to reduce the monitoring frequency of the server; If the operating status coefficient is greater than or equal to the first coefficient threshold, and the operating status coefficient is less than or equal to the second coefficient threshold, it is determined that there is no need to adjust the monitoring frequency of a class of collection servers; if the operating status coefficient is less than the first coefficient threshold, it is determined that the monitoring frequency of a class of collection servers needs to be increased; if the operating status coefficient is greater than the second coefficient threshold, it is determined that the monitoring frequency of a class of collection servers needs to be reduced; When it is determined that the monitoring frequency of a class of collection servers needs to be increased or decreased, dynamic monitoring frequency adjustment is performed according to the adjustment table.

8. The server management method based on the LightGBM model according to claim 7 is characterized in that: For servers in the second category, the monitoring frequency is optimized and adjusted again based on the real-time operation status, including the following steps: Get the operating status coefficients of all servers in the second set. If the operating status coefficient of the server is less than or equal to the second coefficient threshold, the monitoring frequency of the server is not adjusted for secondary optimization. If the operating status coefficient of the server is greater than the second coefficient threshold, the monitoring frequency is adjusted by the operating status coefficient. The adjustment expression is: , where is the monitoring frequency after secondary adjustment, is the monitoring frequency after one adjustment, is the operating status coefficient.

9. The server management method based on the LightGBM model according to claim 8 is characterized in that: Generate corresponding management policies for all servers in the three types of collections, including the following steps: After the management system obtains the information of all servers in the three categories, it marks all servers with a recheck mark and sends it to the administrator; After all servers marked with re-inspection marks have been inspected and repaired, the health status of all servers that have completed inspection and repair is analyzed again based on the LightGBM model; If the server that has been overhauled is classified as a Class II set or a Class I set, the re-inspection mark of the server is deleted and the server is put into use; If the server that has been repaired is still classified as a Class III collection, the server will be deleted from the system library and will no longer be used.

10. A server management system based on the LightGBM model, used to implement the management method according to any one of claims 1 to 9, characterized in that: Including set partitioning module, optimization management module and secondary optimization module; Set division module: When all servers purchased by the enterprise are in trial operation, the trial operation data of all servers are obtained, and after analyzing the health status of all servers based on the LightGBM model, all servers are classified into first-class set, second-class set and third-class set; Optimization management module: Generates corresponding management policies for all servers in the third category set, and optimizes the monitoring frequency of all servers in the second category set based on the health status analysis results; Secondary optimization module: During the use of all servers, each server is monitored at the corresponding monitoring frequency to obtain the real-time operating status of the server. For servers in the first category, the monitoring frequency is dynamically adjusted according to the adjustment table. For servers in the second category, the monitoring frequency is secondarily optimized and adjusted in combination with the real-time operating status. For servers in the third category, corresponding management strategies are generated based on the real-time operating status.

Images